JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 103, NO. B6, PAGES 12,435-12,452, JUNE 10, 1998
Palcomagnetic volcanic data and geometric regularity
of reversals
and
excursions
J. J. Love1
Centre des Faibles Radioactivit6s, Laboratoire Mixte CNRS-CEA, Gif-sur-Yvette, France
Abstract.
Mostly on the basis of paleomagneticsedimentary data, it has been
suggestedthat maps of virtual geomagneticpoles (VGPs), correspondingto
directions of the magnetic field at each site, tend to fall along American and
Asian longitudesduring reversalsand excursions. Such geometric regularity in
transitional fields may indicate that the core and mantle are dynamically coupled.
However, studies of paleomagneticlava data have thus far failed to show any
pattern in transitional fields. In this paper we examine a paleomagneticlava
database coveringreversalsand excursionswhich have occurred over the last 20
Myr. Volcanic eruptions occur sporadically, thus we normalize the data to account
for the fact that
reversal
and excursional
events
at the various
sites are recorded
by different numbersof intermediate directions,but we prefer not to use averaging
methods of previous investigators, who discarded or combined directions when
they appeared to be similar. We find that volcanic data give intermediate VGPs
which tend to fall along American and Asian longitudes, roughly consistentwith
the sedimentary data. This result is not an apparent artifact arising from the
nonuniform geographicdistribution of volcanic sites. Provided the appropriate
polarity is assignedto intermediate VGPs, we find that Icelandic VGPs tend to
fall along Asian longitudes. Other patterns in the data, for example, latitudinal
clustering of VGPs or distinguishinglongitudinal preferencesof excursionsfrom
reversals,are not resolved. However, it appears that in general, transitional fields
are nondipolar.
our understandingof the reversalprocesswill depend,
in large part, on data analysis.
Palcomagnetic studies show that the Earth's magRecently,severalcompilationsof sedimentarypaleonetic field is far from static. It is well known, ini- magneticdata haveappearedto showthat reversalsand
tially deducedfrom a studyof lava bakedclay [Brun- excursionsexhibit geometricregularity; maps of virtual
1.
Introduction
hes, 1906],that the geomagnetic
field occasionally
re- geomagnetic
poles(VGPs),the southmagneticpoleof a
versesits polarity, moreover, inter-reversal periods can dipolecorresponding
to field directionsfrom eachpaleoalso exhibit significantsecular variation with occasional magneticsite, tend to fall alongtwo separatepaths: one
extremedeparturesfrom an axial dipole [Bonhommetalong American longitudesand the other along Asian
and Bahkine,1967],suchexcursions
might themselveslongitudes[Clement,1991; Laj et al., 1991]. These
be considered
abortedreversals[Doelland Cox,1972]. observationshave been greeted with scepticism,with
These phenomena are the most dramatic parts of the
secularvariation of the magnetic field and are the product of convective fluid motion in the Earth's core, the
sourceof the geodynamo. Although theoretical progress
doubt expressedabout the adequacyof the geographic
distributionof sites[Egbert,1992; Valetet al., 1992]
and the reliabilityof sedimentary
records[Barton and
McFadden,1996;Langereiset al., 1992];andarguments
is beingmade [Glatzmaierand Roberts,1995; Holier- over statistics have, thus far, failed to provide convincbachandJones,1993],a completeunderstanding
of the ing resolution[Laj et al., 1992;McFaddenet al., 1993].
geodynamo, and of geomagnetictransitions, remains But the assertionthat intermediateVGPs fall alongsoelusive. What is clear is that further improvementin called preferredlongitudeswas perhapsmost seriously
damagedwhen Prdvot and Camps[1993]publisheda
XNowat Departmentof Earth Sciences,
Universityof Leeds, study of paleomagneticlava records. Lava data, in adUnited Kingdom.
Copyright 1998 by the American GeophysicalUnion.
dition to being independentfrom sedimentarydata, are
not subject to the same controversiesover reliability
which make interpretation of sedimentarydata so dif-
Paper number 97JB03745.
ficult [Hoffmanand Slade,1986]. Pr•vot and Camps
0148-0227/98 / 97JB-03745509.00
found an absenceof preferredVGP longitudesfor geo12,435
12,436
LOVE: GEOMAGNETIC
REGULARITY
magnetic transitions over the last 16 Myr, concluding
that transitional fields are statistically axisymmetric.
The issueof preferredtransitionalVGP longitudesis
important for geodynamotheory; preferred paths may
indicate that the core and mantle are dynamically coupled. Sincethe mantle changesvery slowlycomparedto
the frequencyof reversalsand excursions,the relatively
steady nature of conditions establishedby the mantle
OF REVERSALS
AND EXCURSIONS
and Constable
[1996].We acceptedonlyaveragedirections,inclinations
(I) anddeclinations
(D), takenfrom
at least three magnetically cleaned samplesper flow,
with the precisionparameter c•0s,the semiangleof the
coneof 95% confidence
centeredon the meandirection,
lessthan 20ø. We acceptedonly absoluteintensity mea-
surements(F), when they wereavailable,obtainedby
the Thellier
method.
at the core-mantleboundary(CMB) couldbe responsi- We have taken care to avoidpossibledata overlapby
ble for regularityin transitionalfields[Laj et al., 1991]. requiringthat the data be temporally ordered,coming
Core flow may lock onto mantle heterogeneities,possi- from nonoverlappingstratigraphicsectionsof extruded
bly via topographic
[Hide,1969],electromagnetic
[Run- lava piles. Unfortunately, in practice, it is often difcorn, 1992]or thermal [Bloxhamand Gubbins,1987] ficult to work out the relative stratigraphy; in which
mechanisms,and persistencein coreflow might produce
persistentpatterns in the magneticfield, evenduring reversals and excursions. Of course,theories such as these
are only viable so long as they are motivated and sup-
case, overlap cannot be excludedas a possibility; see,
for exampleAziz ur Rahman[1971,p. 275]. In lieu of
stratigraphy some investigators estimate the temporal
order of paleomagnetic data using radiometric dating
portedby observations;
for a reviewseeGubbins[1994]. of the flows[Abdel-Monemet al., 1972; Mankinen et
In this paper we report on our examinationof paleo- al., 1978;Mankinenet al., 1981],but becauseof uncermagneticlava data coveringreversalsand excursionsfor
the last 20 Myr. Although our database is somewhat
different from that of Pr•vot and Camps, we consider,
for example, only data from stratigraphicsections,the
main differencebetween our analysisand theirs is in the
treatment of the data: we prefer not to averageor discard similar paleomagneticdirections,a practice which,
by itself, can obliterate preferredVGP longitudes.However, because volcanic eruptions occur with extreme
variability we normalize the data to account for the
tainties in the dates, this is often insufficient to insure
fact that the various reversal and excursional
siveigneousbodies(dykes)are of unknowntemporal
events are
recordedat the different sites by widely varying numbers of intermediate directions. We suggestthat our
treatment of paleomagneticlava data is appropriate for
examining the possibility that intermediate VGPs tend
to fall along either American or Asian longitudesor, alternatively, for testing the hypothesisthat the Earth's
transitional magneticfield is statistically axisymmetric.
that unintentional multiple samplingof the same flow
has not occurred and is usually only adequate to allow
approximate temporal ordering. In some cases,investigators, having only partial stratigraphic information,
attempt to place paleomagnetic vectors into temporal
order by finding the smoothest variation in directions
and intensities; see, for example, Roperch and Duncan
[1990].This is subjective,and we find it unacceptable.
We note that paleomagneticvectors taken from intruorder, thus we excluded any data from intrusive bodies.
Finally, if the same site has been visited by more than
one group of investigators, we kept data from the most
detailed study, which was usually the most modern.
We follow Pr•vot and Camps and Valet et al.
[1992]by definingan intermediatedirection(I) as an
inclination-declinationpair with a correspondingVGP
between latitudes +60øN. Other investigators have
2. Method
adopted slightly different definitions, +55øN is common, but clearly, because VGP latitude is a function
2.1.
Data
Selection
of site location and field morphology, defining an inOur databaserelies on publishedrecordsof reversals termedlate direction is somewhat arbitrary; VGP latand excursionswhich occurredduring the last 20 Myr, itudes of the modern field range from about 55øN to
a period of time over which most plate tectonic move- 85øNbecause
the field is not a perfectdipole[Loveand
ment can be neglected relative to the 180ø directional Mazaud,1997];moreover,the fieldduringreversals
and
change of reversals, and a period of time over which excursionsis almost certainly not a perfect dipole eithe conditionsat the CMB have probablynot changed ther. Continuing,a directionis normal (N) if it has
appreciably[McFaddenandMerrill, 1995].The litera- a VGP latitude above60øN and reverse(R) if it hasa
ture surveyof Prgvotand Carups[1993]wasvery thor- VGP latitude below -60øN. The polarity of a transition
ough,and we considereddata from eachof the published
sourcesthat they used;we includeddata from a few papers publishedsincePr4vot and Campsconductedtheir
analysis,and we includeddata from a few other papers
which they did not include. Obviously,we considered
is defined by the relative order of N and R directions; so,
for example,a reversalfromnormal(reverse)to reverse
(normal)is denotedN-R (R-N); whilst a normal (reverse)excursion
is denotedN-N (R-R). Sometimes,
becauseof incomplete recording of a transition in the lava
onlydata setswith intermediatedirections(definedbe- flows,the polarity is undetermined or only partially delow). Our selectioncriteria for eachindividualdatum termined,sofor example,N-I (I-N) denotesa transition
weresimilarto thoseof Camps[1994]and Quidelleur fromnormal(undetermined)to
undetermined
(normal)
et al. [1994]and slightlystricterthan thoseof Johnson final polarity.
LOVE: GEOMAGNETIC
REGULARITY
OF REVERSALS AND EXCURSIONS
12,437
In somecaseswe know the polarity of the transitional direction in our database(when they are not clearly
event becauseit is sui•ciently well dated, for example, determinedby the stratigraphyalone).
the excursional
data from AmsterdamIsland [Watkins
2.2. Data Analysis
Wlno [aj
at.,
the Brunhesepochand thereforerepresentN-N events; Becausevolcanoeserupt sporadically,paleomagnetic
but in most caseswe deducethe polarity of the transitional event from the stratigraphy. In a few cases,
multiple transitions recorded at a particular site are
actually just parts of a single complex transition; in
lava data represent a temporally discontinuousrecord
of geomagneticsecularvariation. It is possiblefor multiple lava flows to be deposited over a duration of time,
short compared to the rate of secular variation of the
which case the transition is counted more than once; magnetic field; in which case the flows will preserve
this is true of the Steens Mountain sequence,for ex- more or less similar records of the magnetic field. In
ample, which is often consideredto be a single reversal an attempt to accountfor such nearly identical data, in
transition[Pr•vot et al., 1985a]but whichmay alsobe
a reversalfollowedby an excursion
[Valetet al., 1985].
The problemhereis that it is oftendifficult(somemight
say "arbitrary"),simplyon the basisof an examination
of a sequenceof paleomagneticlava data, to determine
when one transition beginsand another ends;radiometric dating is inadequate for assigningmost intermediate
data to specificknown events. Therefore we have opted
for simplicity and avoided a great deal of subjectivity
by counting each sequenceof intermediate directions,
bounded by reverse or normal directions, as separate
transitional events. We have found by experiment that
relaxing this strict assignmenthas virtually no effect
on our conclusions;there are few complex transitions
recorded in lava with multiple sequencesof low latitude VGPs. For a minimal amount of resolution, in
part of the analysisbelow we restrict ourselvesto tran-
mostof their analysis,Privot and Camps[1993]combined consecutivedirectionsfrom stratigraphic sections
whenthey appearedto be similar(to within estimatesof
the directionalerrors),and they discardedapparently
redundant directions when they were taken from different neighboring but stratigraphically unconstrained
sites. Such treatments of lava data are fairly common,
McElhinnyet al., [1996]advocatea similarprocedure
which they call "rationalizing" the data, and (•uidelleur
et al. [1994],in their dismissal
of preferredVGP longitudes for inter-transitional data, discardeddata on the
basis of directional similarity.
However, the practice of combining,discarding, and
rationalizing data on the basis of directional similarity is demonstrably dangerous. In Figure 2 we com-
pare, where both directional(I,D) and Thellier intensity measurements
(F) were available,the change
sitionaleventsrecorded(locallyat eachsite) by three in field direction between successive flows • versus the
or more consecutiveintermediatedirections(Valet et change in the field intensity between successiveflows
al., [1988]recommend
four). This alsohasthe advan- AF. Since our database incorporates reversalsand ex-
tage of providing someadditional quality control: single cursionsboth change in direction and change in intendirections are sometimes artifacts due to
sity can be very large. But notice that sometimesthe
laboratory error or overprint.
direction changesby only a few of degrees,an amount
The databaseis summarizedin Table 1 and in Fig- comparable to directional errors, while the intensity
ure 1, wherewe showthe geographicdistribution of sites changesby tens of microteslas,an amount comparable
as well as the geographicdistribution of VGPs per tran- to the present surfaceintensity of the Earth's magnetic
intermediate
sitional
event.
Note that
the distribution
of sites is not
uniform; our database is dominated by data from Ice-
field. Given (say) the presentrate of westwarddrift,
~ 0.2ø/yr, it might (mistakenly)be thoughtthat such
land, and few transitions are recorded in the southern
relatively small differencesin magnetic direction repre-
hemisphere. Furthermore, somesites have many intermediate directions per transitional event, whilst others have few intermediate directionsper event. Among
the various sites, there are 3275 directions, 1318 of
which are intermediate, and 131 absolute Thellier in-
sentonly a few yearsbetweensuccessive
depositions;
on
the other hand, given the rate of decay of the dipole,
~ 15/zT/kyr, the interveningtimesbetweensuccessive
depositionsmay be centuries. Most likely, what is representedhere is a broad range of times, and thus we contensities;507 (local) transitionsare recorded;141 of clude that clustersof VGPs are not necessarilydue exthese(local)transitionsareof knowninitial polarityand clusively to successions
of rapidly depositedlava flows;
are recorded by three or more consecutiveintermedi- they may actually be part of the secularvariation. This
ate directions for a total of 790 intermediate directions.
possibility is supported by measurementsfrom shallow
(The total numberof reversaland excursionalevents marineclay [Valetet al., 1986]andhigh-deposition-rate
oceancores[Channelland Lehman,1997],whichshow
someeventsare recordedat more than one site.) In significantclusteringof intermediateVGPs despitethe
the appendixwe discuss
our reasonsfor excludingparts fact that the vertical accumulation of sediments is much
is smaller than the number of local transitions because
of somedata setsbecausethey comefrom (possibly) more temporally uniform than the vertical accumulaoverlappingneighboringstratigraphicsections,and we tion of lavas. Of course,we are not implyingthat condiscussthe polarity assignmentsto each intermediate secutivepaleomagneticvectors should be combined on
12,438
LOVE: GEOMAGNETIC
ß
ß
i
i
i
i
ß
REGULARITY
i
i
ß
i
i
i
i
ß
OF REVERSALS
ø
ß
ß
AND EXCURSIONS
ß
i
•
i
'
i
LOVE: GEOMAGNETIC REGULARITY OF REVERSALS AND EXCURSIONS
+
+
+ 2 & -, m •= +
• • oo• • +mm
!
!
+ +•
!
!
L=+ + w
!
12,439
12,440
LOVE: GEOMAGNETIC
REGULARITY
gitudes. In the next section we shall demonstrate this
fact using the data themselves.
Clearly, any discarding or combining of field directions should be made independentlyof one's knowledge
of the field directions themselves;to do anything else
introduces bias. Therefore, to test for preferred longitudes, we contend that it is not advisable to combine
and exclude data on the basis of directional similarity.
We recognizethat volcanic eruptions are very sporadic
and someVGP clustersmay not be representativeof the
secular variation. Unfortunately, we have no means of
distinguishinggeomagneticsecularvariation from irreg-
VGPs
Intermediate
OF REVERSALS AND EXCURSIONS
VGPs/Event
ularityof volcaniceruptions(if wedid, wewould).Having said that, in the long run any accidental unevenness
in the distribution of VGPs causedby a few unusually
prolific volcanoesshould be diluted by a preponderance
of data; there is, of course,no reasonto expect that preferred transitional longitudesare the result of volcanoes
preferentially erupting while VGPs happen to lie on one
of the preferred longitudes! Relying on the expectation
that volcanic eruptions and geomagneticvariations are
temporally uncorrelated, we chooseto leave the directions as they are, uncombinedand unrationalized.
Although we do not discardor combinesimilar direc-
Figure1. Geographic
distribution
ofsites,
thesizeof the tionsin an attemptto account
foreruptive
variability
symbol is proportional to the (top) number of intermedi.- during eachtransitional event, we weight the data to acate VGPs and (bottom) number of intermediate VGPs per count for the fact that the total duration of some transitransitional
event.
Note
that
our
database
is dominated
by intermediate directions from Iceland, but a few events;
for example, Steens Mountain and Lousetown in the United
tionsis recorded
at the varioussitesby widelydiffering
States, give many intermediatedirectionsper transition.
•
10 2
--
ß
--
ß
ß
_
the basis of similarity in both direction and intensity;
we present Figure 2 simply to illustrate the fact that
field directions may changemore slowly at some times
than at other times. In other words, directional similarity doesnot necessarilymean near coincidencein time.
We do not a priori know the time dependenceof transitional magnetic fields. It is possiblethat preferred
VPG longitudes, if they are real, may be a property
of the secularvariation whereby field directions change
ß
ß
ß
ß
ß
_
ß
--
ß
II
moreslowly(or arelessscattered)whenVGPs lie along
a preferred longitude and that field directions change
morerapidly(or are morescattered)whenthe VGPsdo
not lie on a preferred longitude. Volcanic data would
preservesuch variations as clusters of VGPs, perhaps
(but not necessarily)
in both latitude and longitude.In
this scenario,if we averagedsimilar directions,then longitudinal preferenceswould be obliterated since VGP
ß
I
10 0
ß
I I I IIIII
10-1
10 o
I
I I I IIIII
I
I I ß1Ilili
101
10 2
Intensity Change IZXFI(•zT)
Figure 2. Directionalchange5 versusintensitychangeAF
between successive lava flows where both directions
and in-
clusters
wouldbe combined
intoa fewor evenone tensity
areavailable.
NotethatAF denotes
a simple
dif-
VGP. We shouldemphasize
oneadditional,but obvious, ference
between
absolute
Thelliermeasurements,
not sim-
point' preferred VGP longitudes may not necessarily pie magnetizationintensities.
Sometimes,
in the intervening
be accompanied
by latitudinalclustering;
intermediatetimebetween
successive
depositions,
thedirection
changes
VGPsmight
bedistributed
over
many
latitudes.
If this by
only
afewdegrees,
while
theintensity
changes
bytens
of
microteslas,an amount comparableto the present surface
isthecase,
thentampering
withthefielddirections
sim- intensity
oftheEarth's
magnetic
field.Thisdemonstrates
ply on the basisof theirsimilarity,
whilstperhaps
not thatthefielddirection
sometimes
changes
verylittleover
completelyremoving,can at least obscurepreferredIon- longperiodsof time.
LOVE: GEOMAGNETIC
REGULARITY
OF REVERSALS
AND EXCURSIONS
12,441
numbersof intermediatedirections.There is precedence to suggestthat in the searchfor patternsfrom onetran-
for suchnormalization:Valet et al. [1992]calculated sitional event to another the data are more readily interthe meanVGP longitude(MVL) for eachtransitionat pretedif the magneticpolelyinginitiallynearthe (say)
eachsite and studied the geographicdistribution of the north geographic
poleis plotted regardlessof magnetic
MVLs (insteadof the VGPs themselves),
thus,in effect, sign; this is what Pr(•vot and Camps call the geomagcounting each transition at each site equally. We find netic convention.
this treatment attractive becauseof the wide range of
One of the advantagesof usingthe geomagneticconfrequencieswith which volcanoeserupt at different sites; vention is that if transitional fields display geometri-
at somesites (for example,Iceland) eachtransitionis cal regularity,perhapsas the result of core-mantlecourecorded by relatively few intermediate VGPs, whilst pling, and if transitionalfieldsare dominantlydipolar,
at othersites(like SteensMountain)a few transitions then VGPs may tend to fall along a single longitude
are recorded by many intermediate VGPs; see Figure
1. Sincewe seekto test the hypothesisthat VGPs from
many transitions tend to fall along preferred longitudes
it is inappropriate to count all VGPs equally; to do
so would bias our study toward just a few transitions
recorded at just a few sites thereby defeating our goal
of finding patterns which might only be evident from a
consideration of many transitions.
In detail the weightingschemewe adopt is a modification of that usedby Valet et al.: we weight eachVGP for
(regardless
of the signof the field);on the otherhand,
for the sanhetransitional fields, usingthe paleomagnetic
conventionwill yield two preferredantipodallongitudes
(depending
on the signof the field). Furthermore,assuminggeometricregularityof transitionalfields,then
under the geomagneticconventionVGPs falling along
(say)twolongitudes
wouldbe evidence
for a nondipolar
transitional field, with the particular longitude being a
functionof sitelocation(underthis conventiona partic-
ular siteshouldgiveVGPsfallingalongonelongitude).
a particulartransitionat a particularsiteby (cos,k)/Nt. Thus, becauseit is perhapsmorephysicallyrelevantand
The cos,Xterm, where ,Xis the latitude of the VGP, gives becauseit might allowfor somediscriminationbetween
more weight to low-latitude VGPs, which is fine, and re- dipolarand nondipolartransitionalfields,we preferthe
moves some of the sensitivity of this analysis to other geomagnetic
convention,and in most of the discussion
definitions of intermediate VGP cutoff. Nx is the total
number
of intermediate
VGPs
for that transition
at that
site, thus dividing by Nx insuresthat each transition at
each site is weighted appropriately. By retaining the
weighted VGPs, instead of the correspondingMVLs,
we get a senseof the geographicdistribution of VGPs,
which is mostly lost when consideringonly mean VGP
longitudes. But for this weighting to be meaningful
we need some minimum
number
of intermediate
which follows we are restricted to data of known initial
polarity. We note that this criterion eliminatesfrom
consideration data from Lousetown Nevada, Necker Island Hawaii, the Oregon Coast and Turkey, as well as
smallpartsof sectionsfrom othersites(whereeventsof
unknowninitial polarityare recorded).
3.
Discussion
VGPs
for each transition. Certainly an averageonly beginsto
3.1. Geographic Distribution of VGPs
havemeaningwith (say)threesamples;
therefore,when
In Figure 3 we plot the (unweighted)intermediate
weightingVGPs by (cos,X)/N•,we consideronly tran- VGPs for reversals and excursions for our database;
sitional events with a minimum
of three intermediate
histogramsshowingtheir longitudinaldistribution usdirections.
We note that this criterion eliminates from
ing both the paleomagneticand geomagneticconyenconsiderationdata from China, Easter Island, France,
New Mexico, New Zealand, R(•union, and Syria, as well
smallpartsof sections
fromothersites(whereeventsare
recordedby only oneor two intermediatedirections).
All
VGPs
Paleo
Conv
2.3. Paleomagnetic and Geomagnetic
Conventions
We concludethis sectionwith a note on VGP plotting
conventions.Many studiesof transitionalfield morphol-
ogyhavedisplayed
VGPs with what Prdvotand Camps
[1993]call the paleomagnetic
convention,
southmagnetic polescorresponding
to field directionsfrom each
paleomagnetic
site plottedon maps.But the equations
of magnetohydrodynamics
(MHD) are invariantunder Figure 3. Map of intermediateVGPs (latitudes between
change
in signof the magnetic
field[Stevenson,
1983], =k60øN)froxnour databaseplottedwith the paleomagnetic
convention(southmagneticpoles). Note that there is much
and therefore,if a reversalinvolvesthe entire magnetic scatter, as well as many clusters of VGPs; the cluster of
field, the senseof the reversal(N-R or R-N) is irrele- VGPs in Africa is from a singlesequenceof lava flowsfor a
vant[Gubbins
andCoe,1993].ThisledHoffman[1984] single transition.
12,442
LOVE: GEOMAGNETIC REGULARITY OF REVERSALS AND EXCURSIONS
Paleo
Convention
Geomag
160
-
160
120
-
120
Convention
Z
80-
8O
40-
40
0
•
0
-180
-90
0
3+I/Event
90
180
•
-90
-180
Paleo Conv
0
3+I/Event
160
-
160
-
120
-
120
-
90
180
Geomag Conv
z
80-
80-
40-
40-
0
•
-180
0
-90
0
90
180
•
-180
Longitude
-90
0
90
180
Longitude
Figure4. Histograms
ofintermediate
(unweighted)
VGPlongitudes
using
both(a,b)thepaleomag-
neticand(c,d)geomagnetic
conventions.
In Figures
4b and4dwehaveonlyincluded
datawhichcomes
fromsequences
of threeor moreconsecutive
interxnediate
directions
pertransitional
event.BothXa
andKolmogorov-Smirnov
(Kuiper)testsconfirmthat theselongitudinal
distributions
of VGPsarenot
uniform,
withsignificance
levels
wellbelow1%.Notethatin Figure4dthereisa hintofa preference
for
VGPs to fall alongAmericanandAsianlongitudes.
tions are shown in Figure 4. The VGPs are not uniformly distributed; there are many clusters of VGPs,
WhenweweightVGPsby (cosA)/N• sothat ourresults are not biasedby a few transitionaleventsfrom
and bothX•' and Kolmogorov-Smirnov
(Kuiper)tests a few siteswherethe volcanoes
just happento have
confirm the obvious observation that the longitudinal eruptedprofusely,then preferredlongitudesbecome
distribution of VGPs is not uniform, with significance clearer.In Figure5 we binnedthe weighted
VGPs in
levelswell below1%. The situationis only slightlydif- latitudeandlongitude
andplottedthemasa gray-scale
ferent when we considertransitional eventsrecordedby
histogram
on a map;sincewe haveadoptedthe geo-
three or more consecutive
magnetic convention,VGPs tend to be concentratedin
intermediate
directions.
Con-
stable[1992],uponanalysisof nontransitional
volcanic the northernhemisphere,the resultof excursionswhich
data, also found a nonuniform distribution of VGP lon- do not alwaysyieldsouthernhemisphere
VGPs. In the
gitudes. And in the one instance when they did not samefigurewealsoshowa histogram
of thelongitudinal
combinesimilar directionsinto singledirections,Prdvot distributionof intermediateweightedVGPs. Note the
and Camps[1993]founda nonuniformdistributionof tendencyfor VGPs to fall alongAmericanand Asian
VGP longitudes, at least for certain tests of nonuni- longitudes
(+75øE);this resultstandsin starkcontrast
formity (Watson's);see their Figure 2b. Pr•vot and to that ofPr•votandCamps,whoreported
nopreferred
Camps found that intermediate VGPs, when they are
countedequally, do not tend to fall alongAmerican and
Asian longitudes. In Figure 4d we seea hint of confinement along those longitudes, but there are numerous
exceptions;for example, the concentration of VGPs in
longitudes.
An interesting
comparison
canbemadenowby plot-
tingVGPsunderthepaleomagnetic
convention;
thehistogramis shownin Figure6. Noticethat althoughthe
histogram
shows
twopreferred
longitudes,
theyareneisouthernAfrica (Figure3) is dueto a singlesequence
of ther as distinctnor are they exactlythe sameas those
lava flows from LousetownNevada; this highlights one in Figure5. The preferred
longitudes
underthe geoof the problemsof not accountingfor the differing num- magnetic
convention
arenot exactlyantipodal;indeed,
ber of intermediate directions per transitional event.
wehaveno reasonto expectthat theyshouldbe. What
LOVE: GEOMAGNETICREGULARITY OF REVERSALSAND EXCURSIONS
12,445
Weighted
/V •Geomag Cony
IllIll
f
..........
,.
_="'
................
30
o
-:30
60
90
0
8
16
24
32
N
16
14
4
2
0 .... I
-180
I .... I
-120
...!
• -60
I
I
I
0
I
60
!
'!
120
!
180
Longitude
Figure 5. Mapof gray-scale
histogram
of intermediate
(weighted,
geomagnetic)
VGPsandhistograms
showing
longitudinal
andlatitudinal
distributions.
Dark(light)shading
indicates
a high(low)average
concentration
of VGPs. All data are weightedby (cos•)/Nz, where• is VGP latitudeand Nz is the
numberof intermediatedirectionsfor eachtransition(Nz _>3). Note a preference
for VGPs to fall along
AmericanandAsianlongitudes;
thereis significant
latitudinalscatter,but thehistogram
doesnot show
anyobvious
tendency
for intermediate
VGPsto clusterat midlatitudes.
In thetop mapthe sizeof the
symbolis proportional
to the contribution
of weightedVGPsfromeachsite.
12,444
LOVE: GEOMAGNETIC REGULARITY OF REVERSALSAND EXCURSIONS
All
Data
Paleo
with VGPs tending to cluster at particular midlatitudes
along Asian and American longitudes. In Figure 5 the
gray-scale histogram shows a good deal of scatter in
the latitudinal distribution of weighted VGPs, which is
almost certainly due to a combination of secular variation and uneven temporal recording of the magnetic
Conv
16
14
,_,12
field due to irregularitiesin volcaniceruptions(recall
that we have no objective way of separating these two
effects).However,the histogramof weightedVGPsdoes
z
not show any obvious tendency for VGPs to cluster at
midlatitudes; thus we are unable to confirm the hypothesis of Hoffman; our observation on this matter is consistent with that of Pr•vot and Camps.
4
2
o
-180
-120
-60
0
60
120
180
Longitude
3.2.
Icelandic
Since •75%
and
Non-Icelandic
Data
of the transitional events in our database
Figure 6. Histogramof intermediate (weighted)VGP lon- come from Iceland it is worthwhile examining the lava
gitudes using the paleoxnagneticconvention.Whilst we see data from this site in some detail.
A measure of the
American and Asian preferred longitudes,we note that tile
paleomagneticconvention (here) does not show them as quality of the Icelandic data can be made by examining
clearly, nor even in exactly the same place, as when VGPs its antipodal symmetry, something which is now recogare plotted under tile geoxnagnetic
convention(Figure 5).
nized as an important test of the fidelity of sedimentary
recordsas well [Clement,1994;Langereiset al., 1992]:
at a particular site, reversedata should have an average
happens,then, whenVGPs are plottedunderthe pa- direction which is opposite that of normal data, an ex-
leomagnetic
convention
is that some(but not all) of pectation which is consistentwith the invariance of the
the VGPswhi& wo,•df•n along(say)+75øE(-75øE) equationsof MHD under changeof sign of the magnetic
under the geomagneticconventionare, instead, plotted
field.
For Icelandic
data we find that
the mean normal
alongthe antipodallongitudeof -105øE (+105øE);this direction(I,D, as•)to be (75.5ø,5.4ø,0.7ø),
whilstthe
has the effect of shifting and broadening the preferred meanreversedirectionis (-75.8 ø, -177.8ø,0.7ø). The
longitudes. Thus one of the advantagesof the geomag- angle between the mean normal direction and the dinetic convention is that it allows for more accurate lorection antipodal to the mean reverse direction is 0.8ø,
cation of nonantipodal preferred VGP longitudes.
which is less than the sum of the a9ss; therefore we
From their study of sedimentary records, Laj et al. conclude that the Icelandic data are statistically an[1992]found somewhatdifferentpreferredlongitudes, tipodally symmetric.
near -60øE and +120øE. The discrepancybetween our
If transitional fields do display geometric regularity
result and that of Laj et al. is probably due to a with preferred VGP longitudes, then we might expect
combinationof scatterin the data (of whichthere is multiple transitions from a single site to give VGPs
much),their consideration
of fewer(local)transitions fallingalonga singlelongitude(solongas the geomagthn we considered
(141), their
of the paleo- netic conventionis adoptedfor plottingVGPs). In Figmagneticconvention(we prefer the geomagnetic
con- ure 7 we plot histograms of the Icelandic intermediate
vention),and of course,there are many technicalproblemswith sedimentaryrecords(mostof whichare irrelevant for lava data). Other comparisons
can be made
with, for example,Clement[1991],who restrictedhis
weighted VGPs using both the paleomagneticand geomagnetic conventions;under the former conventionthe
histogram of intermediate VGPs showstwo longitudinal
peaks, whilst under the later convention the histogram
showsa single longitudinal peak; there is a tendency for
study to the Matuyama-Brunhes reversal,he found preferred VGP longitudesnear -80øE and +100øE; and Icelandic(geomagnetic)
VGPs to fall alongAsian lon-
Coastable[1992],who restrictedherselfto nontransi- gitudes. This observation, of course, does not by itself
tional lava data, found preferred VGP longitudesnear indicate that transitional fields are dipMar; it simply re-90øE and +70øE. Finally, we remark, as others al- flects a pattern in the direction of the transitional field
ready have, that preferred transitional VGP longitudes at Iceland.
are near longitudeswhere magnetic flux is concentrated
If transitional fieldsare nondipolar, then geomagnetic
at the CMB for both paleomagnetictime-averagedmod- VGPs from other sites may follow different paths; this
els [Gabbiasand Kelly, 1993] and historicalmodels point is illustrated in Figure 7 where we see that non[Blozhamand Gubbins,1985],althoughConstablehas Icelandic geomagneticVGPs fall along both American
noted that modern field VGPs are really only clustered and Asian longitudes. Oddly enough, if the paleomagover the American flux patch.
netic convention is adopted, non-Icelandic VGPs tend
Hoffman[1992]hassuggested
that transitionalfields to fall mostly along Asian longitudes; this asymmetry
may, at times, assumean inclined dipole configuration, is almost certainly due to the small number of available
LOVE: GEOMAGNETIC REGULARITY OF REVERSALS AND EXCURSIONS
Iceland
Paleo
Non-Iceland
Conv
Paleo
12,445
Conv
7
14
654-
3-
2-
•
1-
0
0
•
-90
-180
0
Iceland
90
180
-90
0
Non-Iceland
Geornag Conv
90
Geornag
180
Conv
7
14
b
._• 12 -
•
-180
6
d
54o
3
6
[ [_
x
4
z
2
1
0
•
•
-90
-180
0
90
180
Longi•.ude
0
-180
-90
90
180
Longi•.ude
Figure ?. Histogramsof intermediate(weighted)VGP longitudes
for Icelandicand non-Icelandic
data.
For Icelandicdata, under the paleomagnetic
convention,VGPs fall (roughly)on two antipodal longitudes,whilstunderthe geomagnetic
convention
they preferentiallyfall alongone(Asian)longitude.For
non-Icelandic
data the singlepeakseenunderthe paleomagnetic
convention
is split into two (American
and Asian)longitudinalpeaks,an odditywhichwe believeis relatedto the relativelysmallnumberof
non-Icelandic
data.
non-Icelandic data with attendant uneven sampling of
both normal and reverse transitions
at the different
vol-
canic sites. It should, however, be emphasized that
under
neither
convention
are the distributions
of non-
Icelandic weighted VGPs uniform.
3.3.
Subsets
of the
Database
In Figure 8 we showhistogramsof varioussubsetsof
our database,and in Figure 9 we show the corresponding relative contributionto thesehistogramsfrom each
site. Notice that the reversal VGPs by themselvestend
to fall alongAmerican and Asian longitudes,whilst excursionalVGPs tend to fall alongonly Asianlongitudes;
however,both of thesehistogramsshowsignificantscat-
sampling is sparse. Apparent differencesbetween data
which are youngerand data which are older than 10 Myr
are certainly due to differencesin site distribution' the
data from 10-20 Myr come from only two regions, the
western United States and Iceland, and since Icelandic
data give geomagneticVGPs which tend to fall along
Asian longitudesthe histogramof data from 10-20 Myr
also showsan Asian preference.
Next, in Figure 10 we examine the importance of
some of the data selectioncriteria. Up to now we have
considered weighted VGPs coming from transitional
events with three or more intermediate directions, but
notice that we obtain the same preferred American and
Asian longitudesif we considerinstead eventswith four
or more intermediate directions; a similar consistencyis
ter, beingbasedon manyfewerdata than the histogram found if we consider a stricter criterion on each individshownin Figure 5; thus we are hesitant to conclude ual direction, namely four or more magnetically cleaned
that reversals and excursions are somehow fundamen-
samples per flow with a95 < 15ø. Concerning mea-
tally differentfrom eachother. We do not knowthe surement methods, many of the data consideredin this
reasonfor the apparentasymmetrybetweennormaland analysiswereobtainedby alternatingfield (AF) demagreversetransitions;inspectionof the corresponding
site netization of rock samples;however, many researchers
distributionsin Figure 9 showsthat normal interme- prefer thermal (T) measurements.Unfortunately,of
diate data actually comefrom a slightlybetter distri- the sourcepapers consideredhere many authors using
bution of sites, althoughin both casesthe geographic thermal demagnetizationalso used some AF demagne-
12,446
LOVE: GEOMAGNETIC REGULARITY OF REVERSALS AND EXCURSIONS
Reversals
8
-
Normal
Transitions
0-10
8
a
e
0
- 180
- 90
0
90
180
-180
-90
Reverse
Excursions
-
0
-180
•
•
•
-90
•
i
i
0
i
0
Longitude
i
•
90
I
- 180
180
•
- 180
- 90
Transitions
I
•
•
0
10-20
I
I
•
180
90
Myr
d
I
180
90
Myr
I
I
I
I
- 90
I
I
0
I
I
90
Longitude
I
I
I
180
- 180
I
I
- 90
I
I
I
I
0
I
I
90
I
180
Longitude
Figure 8. Histogramsof intermediate (weighted, geoinagnetic)VGP longitudesfor varioussubsetsof
our database. (a) and (b) We coiIlpai'edata for reversalsand excursionsseparately(note that this requires that we know both the initial and final polarities of each transition, a criterion somewhat more
strict than that used in Figure 5, where we only neededthe initial polarity). The reversal data show
somehint of American and Asian preferredlongitudes,but t)oth Figures 8a and 8b also showsignificant
scatter. (c) Normal and (d) reversetransitions both showpreferred Asian longitudes,but only reverse
transitional data show a preferred American longitude. Data from flows (e) youngerthan 10 Myr and
(f) older than 10 Myr show Asian longitudes,but only tile youngerflows show a preferred American
longitude. We believe that tile asymmetry between Figures 8e and 8f is tile result of a poor geographic
distribution of sites and uneven teinporal sampling.
tization but do not clearly state which directions were Camps[1993]by makingthe simpledefinition:for data
obtained with which demagnetizationmethod. When from stratigraphicsections,if the anglebetweensuccesauthorsmentionboth demagnetizationmethods,we as- sivedirections5 is lessthan the sumof the correspondsume that AF results are consistent with thermal reing c•s, then these two directions are considered to
sults; the subset of our database which comesfrom these belongto the samedirectionalgroup;this groupingis
papers is very small, and yet, it showssome Ameri- applied along the stratigraphic sectionuntil 5 exceeds
can and Asian longitudinalconfinementof intermediate the c• cutoff. With this definition, singledirections
VGPs. Finally, some of the data in our database were can be separatedirectional groups,but often, multiple
obtainedby eitherblanketdemagnetization
(usuallyat directionsfall into a singlegroup;examinationof spe15 mT) or simpletwo-stepdemagnetization
(usuallyat cificsequences
of data showsthat this simpledefinition
10 and 20 mT); many Icelandicdata fall into this cate- givesdirectionalgroupslike thoseof Pr6.votand Camps.
gory. Although blanket demagnetizationmay at times
Repeatingollr analysison thesedirectionalgroups,
be acceptable, often only 10 or 15 mT are sufficient we obtain the histograms in Figure 11. Note that
to removeoverprintfrom Icelandiclavas[Kristjdnsson applyingdirectionalgroupingreducessignificantlythe
andYdhannesson,
1996];if we consideronlydata which number of directions;for the unweightedintermediate
has been demagnetizedwith three or more steps, we (grouped)VGPs (Figure11a) a X2 test indicatesthat
find that this too yields preferred American and Asian the longitudinal distribution of VGPs has a 60% chance
longitudes for intermediate VGPs. Taken as a whole, of being drawn from a uniform population;the reader
Figure 10 indicates that our conclusionsare relatively will recall that without groupingthis probability was
insensitiveto more stringent data selectioncriteria.
lessthan 1%. Furthermore,whenwecomparethe distri-
butionof weightedgroupedVGPs (Figure11b)with the
weightedbut ungroupedVGPs in Figure 5, we seethat
3.4. On Grouping Similar Directions
grouping data by directional similarity all but erases
Toward examining the effect of combining directions the preferredAmericanand Asian longitudes,which in
on the basis of their similarity, we follow Privot and hindsightis hardly surprising.
LOVE: GEOMAGNETIC REGULARITY OF REVERSALS AND EXCURSIONS
Reversals
Normal
Transitions
Reverse
Transitions
0-10
12,447
Myr
Figure 9. Maps, corresponding
to the histogramssitownin Figure 8, witerethe sizeof symbolis proportionalto the contrit,utionof VGPs frowneachsite weightedby (cosA)/N•, witereA is VGP latitude
and N• is the number of intermediate directionsfor each transition (N• > 3).
4+I/Event
Thermal
Demag
4
12
3
x
4--•
z
2
0
If
•
-180
-90
0
90
180
-180
a95<15ø 4+ Samples/Flow
•_6
-90
0
3+Step
b
90
180
Demag
B
0
-180
-90
0
Longitude
90
180
-180
-90
0
90
180
Longitude
Figure 10. Histogra•ns
ofintermediate
(weighted,
geomagnetic)
VGPlongitudes
fordifferent
(stricter)
data selection
criteria. (a) We plot VGP longitudes
fromeventswitIt N• > 4; (b) data with four or
moremagnetically
cleaned
samples
per flowandwith ct95< 15ø;(c) the (few)datafromsiteswhere
samples
weresubjected
to at leastsomethermalde•nagnetization'
(d) •neasurements
madewith at least
threestepsin step-wise
demagnetization.
Takenasa whole,thesehistograms
indicatethat the apparent
preference
for transitional
VGPsto fall alongAmericanandAsianlongitudes
is relativelyinsensitive
to
the data selection
criteria.
12,448
LOVE: GEOMAGNETIC
Grouping
Similar
REGULARITY
Directions
50
4030Z
2010-
OF REVERSALS AND EXCURSIONS
canicsites,but wedidnot combinedirections
simplyon
the basisof their similarity.We appreciatethe motivation that othershavehad whencombiningor discarding similarpaleomagnetic
directions.Sporadicvolcanic
eruptionscangiveVGP clustering;
however,changes
in
the secularvariationitself can also give VGP clustering. Unfortunately,there is no objectivemeansof untanglingthesetwo effects.With enoughdata it should
not be necessary;
obviously,the frequencyof volcanic
eruptions are unrelated to, and therefore uncorrelated
0
•
-180
-90
0
90
180
with, variationsin the magneticfield. Barringa remarkablecoincidence,
thereis no reasonto expectthat
preferredVGP paths are the result of volcanicvariabil-
16
ity. Thus, in the searchfor preferredVGP pathsit is
difficultto justifycombining
or discardingsimilardirections; more importantly,the practiceitself introduces
bias. SinceVGP pathsare (at eachsite)equivalent
to
themagnetic
fieldpointingpreferentially
withina range
of certaindirections,tamperingwith similardirections
1 4208-
6I
will changethe geographicdistribution of VGPs. More
specifically,
combiningor discardingsimilardirections
will, inevitably,homogenize
the geographic
distribution
4Z
2-
0
of VGPs.
-180
-90
0
90
180
Longitude
In no case,from our analysisof paleomagnetic
lava
data, havewe founduniformlongitudinaldistributions
Figure11. Histograms
ofintermediate
(geomagnetic)
VGP of intermediate
VGPs;therefore
weconclude
thattranlongitudes
for directionalgroups:(a) unweighted
VGPs and
(b) weightedVGPs, in both casesfor eventswith N• > 3.
Combiningdata into directionalgroupsreducesthe number
of transitional events to 86 and the number of intermediate
directionsto 403, comparedto 141 and 790, respectively,
for ungrouped
data. For Figure11aa X2 testindicatesthat
the distributionhas a 60% chanceof beingdrawnfrom a
uniformpopulation. Clearly,combiningdata on the basis
of directionalsimilarityeffectively
hoxnogenizes
the longitudinal distribution:compareFigurelib with Figure5.
3.5.
On
Site
Common
120
100
Site Long
a
80
Bias
Despitethe relative reliability of paleomagneticlava
data, preferredpaths may be an artifact of a poor ge-
0
-180
ographic distribution of sites combined with a bias inherent to VGP mapping whereby random noisecauses
VGPs to fall alonglongitudes+90øE from site longitudes[Egbert,1992].To checkfor sitedistributionbias,
we considerVGP longitude minus site longitude, or
commonsite longitude;seeFigure 12. Sincewe do not
see peaks at both -90øE and +90øE we conclude that
-90
0
90
180
16
14-
z
12-
,< 10•
8-
•x
6-
0
[I
the biassuggested
by Egbert is not an important factor
4in the apparent longitudinal preferenceof intermediZ
2ate VGPs foundhere. The tendencyfor the weighted
0
VGPs to fall +90øE of the site longitudeis due to Ice-180
-90
0
90
180
landicdata (Icelandhappensto sit on a longitudeabout
-90øE of the Asian longitude, where most of the IceVGP Long - Site Long
landicVGPs fall).
Figure 12. Histogramsof (geomagnetic)VGP common
site longitude(VGP longitudeminus site longitude): (a)
4.
Conclusions
unweightedVGPs and (b) weighted VGPs, in both cases
for events with N• > 3. In neither case do the VGPs tend
In this analysiswe useda weighting method which ac- to fall alonglongitudes4-90ø from site longitude;therefore
number of intermediate
VGPs
the preferredlongitudesseenin Figure 5 are not an artifact
recording each transitional event at the different vol- of site bias like that suggested
by Egbert[1992].
counts for the different
LOVE: GEOMAGNETIC
REGULARITY
sittonal fields are not statistically axisymmetric. More
specifically,we have found evidence,as others have from
sedimentarydata, that the geometry of the geomagnetic
field during transitions appears to repeat sufficiently
often over millions of years to give preferred American and Asian VGP longitudes. Interestingly, for Ice-
OF REVERSALS AND EXCURSIONS
12,449
Iceland, Esja, KrisOdnssonet al. [1980]: Keep nonoverlappingcompositesectionFA02-SC1 specifiedoil page 38.
Iceland, TrSllaskagi,Saemundssonet al. [1980]: Keep
nonoverlapping composite section PA02-PG60 specified on
page12. PA02-PC39(PC40-PG60) are older(younger)than
10 Myr; scc Figure 4.
Iceland, SK1-JF159, SR0-BX08, BT1-BV55, McDougall
land alone,underthe geomagnetic
convention,
VGPs etal. [1984]:
Kccpnonoverlapping
composite
sections
speci-
tendto fall alongAsianlongitudes;
thusthe transi-fiedinFigure2.
SR0-BD10
(BD12-BX08)
areolder
(younger)
tional field displays regularity at a single site. Since
than 10 Myr; seeFigure 7.
Iceland,Tj•rnes, Kristjdnssonet al. [1988]:Possibleover-
the VGP depends
on siteposition
andfieldmorphol-lapwithGF,BA,GM;keep
onlyGS.
ogythe presence
of twopreferredVGP longitudes
(ge-
Iceland,Langidalur,
K'ristjdnsson
et al. [1992]:Keep
omagneticconvention)from data from differentsites nonoverlapping
coInposite
sectionTL1-IN85 specified
oil
indicates
thattransitional
fieldsarenondipolar.
Thesepage
39.TL1-33
wcrcdeInagnctizcd
withtwosteps;
thereInaining flows wcrc dcInagnetizedwith three or Inore steps:
observationsare contradictory to those of Pre'vot and see page 38.
'
Camps
[1993].
Finally,
latitudinal
clustering
oftransi- Iceland,
R3-N3,
Kristjdnsson
andSigurgeirsson
[1993]'
tionalVGPs[Hoffman,
1992]issimplynotsupported
by Possible
overlap
withMO,KY, SE,FL,BO,MU;keeponly
our (large)database.Becauseof the limitedqualityand HU, whichis R-N.
numberof data,additionalconclusions
remainelusive; Iceland,Mj6ifjSr0ur,
Kristjdnsson
eta[. [1995]:Keep
wewere
unable,
forexample,
tocome
toanyconclusion
nonoverlapping
composite
section
DA1-MC56
specified
on
regarding
differences
(orsimilarities)
inVGPdistribupage
824.•safja0ardjfip,Kristjdnssonand Jdhannesson
Iceland,
tionbetween
reversals
andexcursions.
Clearlycontin- [1996]:
Keepnonoverlapping
composite
section
DO1-DM14
ued progresswill benefit from the ongoingcollectionof specified
on page12.
paleomagnetic
lava data.
Nevada,Lousetown,
Heinrichs[1967]:Possibleoverlap
Appendix:Noteson Data Sets
with Clark sectionand other nonstratigraphicdata; keepF1F34. Originally thought to be R-N, but recently Robertsand
Shaw
[1990]
haveconcluded
thatflowF1isnotR;therefore
Mostoftheusefulinformation
aboutourdatabase
issum- Lousetown
sequence
is I-N.
marized
inTable1. Herewemention
some
technical
points Oregon,
Steens
Mountain,
Mankinen
etal.[1985],
Prdvot
relating
topossible
overlap
ofneighboring
stratigraphic
sec-etal.[1985b]'
Possible
overlap
withsections
B andC'keep
tions,
some
issues
related
to dating
anddemagnetization,
section
A. Measurements
fromflows
A41andA42appear
andassignment
oftransitional
polarity.
Usually,
thepolar-torecord
extremely
rapidgeomagnetic
variations
over
their
ityoftheintermediate
dataaredetermined
bythestratig-thicknesses,
possibly
during
theirthermal
cooling,
butthis
raphy,
butin some
cases,
where
thestratigraphy
byitselfrectional
iscontroversial;
omitmeasurements
fromA41andA42.Digroupsnot accepted.
does not fix the polarity, the polarity is determined either
bytheageofthedataorbyothergeological
informationPolynesia,
Huahine,
Roperch
andDuncan
[1990]'
Some
taken from the sourcepapers.
data do not come from stratigraphic sections. No clear
Alaska,
WaitCreek,andCastles,
Bingham
andStonestratigraphic
relationship
between
smallsections
could
be
[1976]'
W andC areN-N.
worked
out(see
page
2714).Onpage
2722
authors
offer
a
Amsterdam
Island,
Watkins
andNougier
[1973]'
Onlygrand
composite
section
ofthedatainTable
2,andTable
5
flows
17,18,19arefromstratigraphic
section.
Amsterdam
summarizes
thissection
interms
ofdirectional
groups.
To
Island
intermediate
dataareBrunhes
ageandtherefore
are avoid
subjective
interpretation
ofthepaleomagnetic
record,
N-N.
aswellaspotentialoverlaps
in stratigraphic
sections;
keep
Arizona,
Hackberry
Mountain,
McKee
andElston
[1980]'only83H21-D40
andF66-68.
Intermediate
dataareN-I.
Omitallsedimentary
data.B27b-B9
(B8-T1)
areolder Sicily,
Vulcano,
Lajetal.[1997]'
Initialpolarity
ofVul(younger)
than10Myr;see
Figure
2.
cano
intermediate
datawas
notmeasured,
butintermediate
Canary,
LaPalma,
Quidelleur
andValet
[1996]'
Possible
dataareBrunhes
ageandareN-N.
overlap
withLS;keep
LL.
Syria,
Levant,
Roperch
andBonhommet
[1986]:
Possible
Easter
Island,
Isaacson
andHeinrichs
[1976]'
Onlyflowsoverlap
among
different
small
sections;
keep
S14-S19.
32,33,34,which
come
froma stratigraphic
section,
are Tahiti,
Punaruu
Valley,
Chauvin
etal. [1990]:
Onthe
considered;
ofthese,
34hasc•a• 20ø.Easter
Island
inter-basis
ofradiometric
dating,
BKH-BKL
taken
aslower
Cobb
mediate
dataareBrunhes
ageandtherefore
areN-N.
Mountain
(R-N)transition,
whilst
BKN-BKT2
aretaken
as
France,
Laschamp,
Roperch
etal. [1988],
Chauvin
etal. upper
Cobb
Mountain
(N-R)transition.
Remaining
polarit1989]'
Samples
taken
froInthree
separate
flows.
Laschamp
tiesofPunaruu
datadetermined
bystratigraphic
section.
intermediate
dataareBrunhes
ageandtherefore
areN-N. Washington,
Grande
Ronde,
Choiniere
andSwanson
Georgia,
Akhalkalaki,
Camps
etal.[1996]'
Possible
over-[1979]'
Possible
overlap
withGRofHooper
etal.[1979].
lapwithX,Y,Z,keep
W. Initialpolarity
ofdatawasnot Washington,
Saddle
Mountain,
Choiniere
andSwanson
measured,
butageofflows
corresponds
toareverse
Gilbert[1979]'
OI1-EM1
(BC3-LM1)
areolder
(younger)
than10
epoch;
thereibre
intermediate
dataareR-R.
Myr;seeFigure
2 ofHooper
etal. [1979].
Iceland,JSkuldalur, Watkinset al. [1975]:Possibleoverlap with GK, GJ; keep GH1-GL5.
Acknowledgments. We thank L. Kristj•inssonfor proIceland, Bessastadaa,McDougall et al. [1976]' Possible viding most of the Icelandic data used in this analysis.
overlapwith EW; keep EQ.
We thank P. Camps, D. Gubbins, C. Kissel, C. Laj, A.
Iceland, BorgarfjSrdur, Watkins et al. [1977]: Keep Mazaud, M. Pr•vot for useful discussions.We thank the
nonoverlappingcompositesection NP1-NT 112 specifiedin AssociateEditor, A. Jackson,and C. Constableand AnonyFigure 2 of McDougall et al. [1977].
mous for providing critical reviews of this paper. This
12,450
LOVE: GEOMAGNETIC
REGULARITY
work was supported by the Leverhulme Trust and a "Bourse
Chhteaubriand" from the French government.
References
Abdel-Monem, A., N. D. Watkins, and P. W. Gast,
Potassium-argonages,volcanicstratigraphy and geomagnetic polarity history of the Canary Islands: Tenerife, La
Palma, and Hierro, Am. J. $ci., 272,805-825, 1972.
Aziz ur Rahman, Palaeomagneticsecularvariation for recent
normal and reversedepochs,from the newer volcanicsof
Victoria, Australia, Geophys.J. R. Astron. Soc., 24,255269, 1971.
Barton, C. E., and P. L. McFadden, Inclination shallowing
and preferred transitional VGP paths, Earth Planet. Sci.
Left., 140, 147-157, 1996.
Bingham, D. K., and D. B. Stone, Palaeosecularvariation
OF REVERSALS AND EXCURSIONS
ing the Matuyama-Brunhes geomagneticreversal, Earth
Planet. Sci. Left., 104, 48-58, 1991.
Clement, B. M., Geomagneticpolarity transition records
from ODP cores,Joint Oceanogr.Inst.; U.S. Sci. Advis.
Comm. Workshop Report, Fla. International Univ., Miami, 1994.
Coe, R. S., S. Gromm6, and E. A. Mankinen, Geomagnetic palcointensitiesfrom excursion sequencesin lavas
on Oahu, Hawaii, J. Geophys.Res., 89, 1059-1069, 1984.
Constable,C. G., Link betweengeomagneticreversalpaths
and secular variation of the field over the past 5 My,
Nature, 358,230-233, 1992.
Dagley, P., and E. Lawley, Palaeomangeticevidencefor the
transitional behaviourof the geomagneticfield, Geophys.
J. R. Astron. Soc., 36, 577-598, 1974.
Dagley, P., R. L. Wilson, J. M. Ade-Hall, G. P. L. Walker,
S. E. Haggerry, T. Sigurgeirsson,N. D. Watkins, P. J.
of the geomagneticfield in the Aleutian Islands, Alaska,
Smith, J. Edwards, and R. L. Grasty, GeomagneticpoGeophys. J. R. Astron. Soc., 28,317-335, 1972.
larity zonesfor Icelandiclavas,Nature, 216, 25-29, 1967.
Bingham, D. K., and D. B. Stone, Evidencefor geomagnetic Doell, R. R., Paleomagnetismof volcanicrocksfrom Niihau,
field excursionsand secular variation from the Wrangell
Nihoa, and NeckerIslands,Hawaii, J. Geophys.Res., 77,
volcanicsof Alaska, Can. J. Earth Sci., 13,547-554,1976.
Bloxham, J., and D. Gubbins, The secular variation of the
Earth's magnetic field, Nature, 317, 777-781, 1985.
Bloxham, J., and D. Gubbins, Thermal core-mantle interactions, Nature, 325,511-513, 1987.
3725-3730, 1972.
Doell, R. R., and A. V. Cox,The Pacificgeomagnetic
secular
variation anomaly and the questionof lateral uniformity
in the lower mantle, in The Nature of the Solid Earth,
edited by E. C. Robertson, pp. 245-284, McGraw-Hill,
Bogue, S. W., and R. S. Coe, Transitional palcointensities
New York, 1972.
from Kauai, Hawaii, and geomagneticreversalmodels,J. Doell, R. R., and G. B. Dairytopic, Potassium-argonages
Geophys. Res., 89, 10341-10354, 1984.
and palcomagnetismof the Waianae and Koolau volcanic
Bonhommet, N., and J. Babkine, Sur la pr6sencede direcseries, Oahu, Hawaii, Bull. Geol. Soc. Am., 84, 1217-
tionsinvers6es
dansla Cha]nedesPuys,C. R. Acad.Sci.
Paris, 264, 92-94, 1967.
Brown, L., J. Pickens, and B. Singer, Matuyama-Brunhes
transition
recorded
in lava flows of the Chilean
Andes:
Evidence for dipolar fields during reversals,Geology, 22,
299-302, 1994.
Brunhes, B., R6cherchessur le directions d'aimantation des
roches volcaniques, J. Phys., 5,705-724, 1906.
Camps, P., Comportment du champ magn6tique de la terre
au cours de sesrenversements:Etude d'un exemple, variations ultra rapides et caract6ristiquesstatistiques globales, Ph.D. Thesis, Univ. of Montpellier, Montpellier,
1994.
1242, 1973.
Doell, R. R., G. B. Dalrymple, R. L. Smith, and R. A. Bailey,
Palcomagnetism,potassiuxn-argon
ages,and geologyof
rhyolitesand associatedrocksof the Vailes Caldera, New
Mexico, Mere. Geol. Soc. Am., 116, 211-248, 1969.
Egbert, G. D., Samplingbias in VGP longitudes,Geophys.
Res. Left., 19, 2353-2356, 1992.
Glatzmaier, G. A., and P. H. Roberts, A three-dimensional
self-consistent
computersimulationof a geomagneticfield
reversal, Nature, 377, 203-209, 1995.
Groinre6,C. S., T. A. Reilly, A. E. Musseft,and R. L. Hay,
Palaeomagnetism
and potassium-argon
agesof volcanic
rocksof NgorongoroCaldera, Tanzania, Geophys.J. R.
Camps, P., G. Ruffet, V. P. Shcherbakov, V. 'V.
Astron. Soc., 22, 101-115, 1970.
Shcherbakova, M. Pr6vot, A. Moussine-Pouchkine, L.
Gubbins, D., Geomagneticpolarity reversals:A connection
Sholpo,A. Goguitchaichvili,and B. Asanidz6,Paleomagwith secularvariationand core-mantleinteraction?,Rev.
netic and geochronologicalstudy of a geomagneticfiehi
Geophys., 32, 61-83, 1994.
reversal or excursion recorded in pliocene rocks from
Gubbins,
D., and R. Coe, Longitudinally confined geomagGeorgia(LesserCaucasus),Phys. Earth Planet. Int., 96,
netic
reversal
paths from non-dipolar transition fields,
41-59, 1996.
Nature, 362, 51-53, 1993.
Channell, J. E. T., and B. Lehman, The last two geomag-
netic polarity reversalsrecorded in high-deposition-rate Gubbins,D., and P. Kelly, Persistentpatterns in the geoxnagneticfield during the last 2.5 Myr, Nature, 365,
sediment drifts, Nature, 389, 712-715, 1997.
829-832, 1993.
Chauvin, A., R. A. Duncan, N. Bonhommet, and S. Levi,
Hamilton, T. S., and M. E. Evans,A magnetostratigraphic
Palcointensity of the Earth's magnetic field and K-Ar
and secularvariation study of Level Mountain, northern
dating of the Louchadii•revolcanicflow (central France):
British Columbia, Geophys.J. R. Astron. Soc., 73, 39-49,
New evidencefor the Laschamp excursion, Geophys.Res.
Left., 16, 1189-1192, 1989.
Chauvin, A., P. Roperch, and R. Duncan, Records of geomagnetic reversalsfrom volcanic islands of French Polynesia 2. Palcomagneticstudy of a flow sequence(1.2-0.6
Ma) from the island of Tahiti and discussionof reversal
models, J. Geophys.Res., 95, 2727-2752, 1990.
Choiniere, S. R., and D. A. Swanson,Magnetostratigraphy
and correlationof miocenebasaltsof the northern Oregon
coast and Columbia Plateau, southeastWashington,Am.
J. Sci., 279, 755-777, 1979.
Clement, B. M., Geographical distribution of transitional
VGP's: Evidence for non-zonal equatorial symmetry dur-
1983.
Heinrichs, D. F., Palcomagnetismof the Plio-Pleistocene
Lousetownformation,Virginia City, Nevada,J. Geophys.
Res., 72, 3277-3294, 1967.
Hide, R., Interactions betweenthe Earth's liquid core and
solid mantle, Nature, 222, 1055-1056, 1969.
Hoffman,K. A., Dipolar reversalstate of the geomagnetic
field and core-mantledynamics,Nature, 359, 789-794,
1992.
Hoffman,K. A., Geoxnagnetic
reversals,
Nature, 309, 90-91,
1984.
Hoffman,K. A., and S. B. Slade,Polarity transition records
LOVE: GEOMAGNETIC REGULARITY OF REVERSALS AND EXCURSIONS
12,451
and the acquisition of remanence: A cautionary note,
Geophys. Res. Left., 13,483-486, 1986.
Holierbach,R., and C. A. Jones,Influenceof the Earth's in-
Mankinen, E. A., J. M. Donnelly, C. S. Gromm•, and B.C. J.
Hearn, Paleomagnetismof the Clear Lake volcanicsand
new limits on the age of the Jaramillo normal-polarity
event, U.S. Geol. Surv. Proœ.Pap., 1141, 67-82, 1981.
ner core on geomagneticfluctuationsand reversals,Nature, 365,541-543, 1993.
Mankinen, E. A., M. Pr•vot, C. S. Gromm•, and R. S.
Hooper, P. R., C. R. Knowles, and N. D. Watkins, MagneCoe, The SteensMountain (Oregon)geomagneticpolarity transition 1. Directional history, duration of episodes,
tostratigraphyof the Imnaha and Grande Ronde basalts
in the southeast part of the Columbia Plateau, Am. J.
and rock magnetism, J. Geophys.Res., 90, 10393-10416,
1985.
Sci., 279, 737-754, 1979.
Isaacson, L. B., and D. F. Heinrichs, Paleomagnetism and McDougall, I., and N. D. Watkins, Age and duration of the
secular variation of Easter Island basalts, J. Geophys.
R•union geomagneticpolarity event, Earth Planet. Sci.
Left., 19,443-452, 1973.
Res., $1,1476-1482, 1976.
Johnson,C. L., and C. G. Constable,Palaeosecularvariation McDougall, I., L. KristjKnsson,and K. Saemundsson,Magrecorded by lava flows over the past five million years,
netostratigraphy and geochronologyof northwest IcePhilos. Trans. R. Soc. London Ser. A, 354, 89-141, 1996.
land, J. Geophys. Res., $9, 7029-7060, 1984.
Korio, M., Intensity of the Earth's magneticfield in geologi- McDougall, I., K. Saernundsson,H. Johannesson, N. D.
cal time III. Pleistocene and Pliocene data from Japanese
Watkins, and L. KristjKnsson,Extension of the geomagvolcanicrocks, J. Geornagn.Geoelectr.,23, 1-9, 1971.
netic polarity time scaleto 6.5 m.y.: K-At dating, geologKono, M., Paleoamagnetismof PleistoceneUsami Volcano,
ical and paleoxnagneticstudy of a 3,500-m lava succession
Izu Penninsula, Japan-Intensity of the Earth's magnetic
in western Iceland, Bull. Geol. Soc. Am., $$, 1-15, 1977.
field in geologicaltime II, J. Geomagn. Geoelectr., 20, McDougall, I., N. D. Watkins, and L. KristjKnsson,
353-366, 1968.
Geochronology and paleomagnetism of a MioceneKristjKnsson,L., New palaeomagneticresultsfroxnIcelandic
Pliocene lava sequenceat Bessastadaa, eastern Iceland,
Neogenelavas, Geophys.J. Int., 121,435-443, 1995.
Am. J. Sci., 276, 1078-1095, 1976.
Kristj&nsson,L., and H. J6hannesson,Stratigraphy and pa- McElhinny, M. W., P. L. McFadden, and R. T. Merrill, The
myth of the Pacific dipole window, Earth Planet. Sci.
leomagnetism
ofthelavapilesouth
of•safja0ardjfp,
NWLeft., 143, 13-33, 1996.
Iceland, J6kull, 44, 3-16, 1996.
KristjKnsson,L., and M. Sigurgeirsson,The R3-N3 and R5- McFadden, P. L., C. E. Barton, and R. T. Merrill, Do virtual
geomagneticpolesfollow preferred paths during geomagN5 palaeomagnetictransition zonesin SW-Iceland revisnetic reversals?,Nature, 361,342-344, 1993.
ited, J. Geomagn. Geoelectr., 45,275-288, 1993.
Kristj•nsson, L., I. B. Fridleifsson, and N. D. Watkins, McFadden, P. L., and R. T. Merrill, History of Earth's
Stratigraphy and paleomagnetismof the Esja, Eyrarfjall
magnetic field and possible connections to core-mantle
and Akrafjall Mountains, SW-Iceland, J. Geophys., 47,
boundary processes, J. Geophys. Res., 100, 307-316,
31-42, 1980.
KristjKnsson, L., A. Gudmundsson, and H. Hataidsson, Stratigraphy and paleoxnagnetism
of a 3-km-thick
Miocene lava pile in the Mj6ifjSrdur area, eastern Iceland, Geol. Rundsch., 84,813-830, 1995.
1995.
McKee, E. H., and D. P. Elston, Reversalchronologyfrom a
7.9- to 11.5-M.Y.-old volcanic sequence in central Arizona: Coxnparisonwith ocean floor polarity record, J.
Geophys. Res., $5,327-337, 1980.
Kristjfmsson,
L., H. J6hannesson,
J. Eir•ksson,
and A. I. Patel, J.P., and P. K. S. Raja, Paleomagneticresultsfrom
Gudmundsson, Brunhes-Matuyama paleomagnetism in
the Narosura and Magadi volcanics of Kenya, Phys.
three lava sections in Iceland, Can. J. Earth Sci., 25,
Earth Planet. Int., 19, P7-P14, 1979.
215-225, 1988.
Pr•vot, M., and P. Camps,Absenceof preferredlongitudinal
KristjKnsson,L., H. J6hannesson,and I. McDougall, Stratigsectorsfor poles from volcanic records of geomagnetic
reversals, Nature, 366, 53-57, 1993.
raphy, age and paleomagnetismof Langidalur, northern
Iceland, JiSkull, 42, 31-44, 1992.
Pr•vot, M., E. A. Mankinen, R. S. Coe, and C. S. Gromm•,
Laj, C., A. Mazaud, R. Weeks, M. Fuller, and E. HerreroHow the geomagneticfield vector reversespolarity, NaBervera, Geomagneticreversalpaths, Nature, 351,447,
ture, 316, 230-234, 1985a.
1991.
Pr•vot, M., E. A. Mankinen, R. S. Coe, and C. S. Gromm•,
Laj, C., A. Mazaud, R. Weeks, M. Fuller, and E. HerreroThe Steens Mountain (Oregon) geomagneticpolarity
Bervera, Statistical assessmentof the preferred longitutransition 2. Field intensity variations and discussion
dinal bandsfor recentgeomagneticreversalrecords,Geoof reversal models, J. Geophys. Res., 90, 10417-10448,
1985b.
phys. Res. Left., 19, 2003-2006, 1992.
Laj, C., A. Ra'•s,J. Surmont,P. Y. Gillot,H. Guillou,C. Quidelleur, X., and J.P. Valet, Geomagneticchangesacross
the last reversal recorded in lava flows from La Palma,
Kissel, and E. Zanella, Changesof the geomagneticfield
vector obtained from lava sequenceson the island of VulCanary Islands, J. Geophys. Res., 101, 13755-13773,
1996.
cano (Aeolian Islands, Sicily), Phys. Earth Planet. Int.,
99, 161-177, 1997.
Quidelleur, X., J.P. Valet, V. Courtillot, and G. Hulot,
Langereis,C. G., A. A.M. V. Hoof, and P. Rochette, LonLong-term geometry of the geomagneticfield for the last
gitudinal confinement of geomagneticreversal paths as
5 million years: An updated secular variation database
a possible sedimentary artefact, Nature, 358, 226-230,
from volcanic sequences,Geophys. Res. Left., 21, 16391992.
1642, 1994.
Love, J. J., and A. Mazaud, A database for the Matuyama- Roberts, N., and M. Fuller, Similarity of new palaeomagnetic
data from Santa Rosa Mountains
with those from Steens
Brunhes magnetic reversal, Phys. Earth Planet. Int., 103,
207-245, 1997.
Mountain gives wide regional evidence for a two-stage
Mankinen, E. A., J. M. Donnelly, and C. S. Gromm•, Geoprocess of geomagnetic field reversal, Geophys. J. Int.,
100, 521-526, 1990.
magnetic polarity event recordedat 1.1 m.y.B.P. on Cobb
Mountain, Clear Lake volcanicfield, California, Geology, Roberts, N., and J. Shaw, Secondary magnetizations cast
6, 653-656, 1978.
doubt on exceptional palaeointensity values from the
12,452
LOVE: GEOMAGNETIC REGULARITY OF REVERSALS AND EXCURSIONS
LousetownCreek lavas of Nevada, USA, Geophys.J. Int.,
101,251-257, 1990.
Roperch, P., and N. Bonhommet,Paleomagnetismof
Miocene volcanism from soutit Syria, J. Geophys., 59,
98-102, 1986.
lava sectionsin southwesternIceland, J. Geomagn. Geoelectr., 47, 89-102, 1995.
Valet, J.P., C. Laj, and P. Tucholka, High-resolution sedimentary record of a geomagneticreversal, Nature, 322,
27-32, 1986.
Roperch,P., and R. A. Duncan, Recordsof geomagnetic Valet, J.P., C. Laj, and P. Tucholka, Volcanic record of
reversalsfrom volcanic islands of French Polynesia 1. Pa-
reversal, Nature, 316, 217-218, 1985.
leomagneticstudy of a polarity transitionin a lava se- Valet, J.P., L. Tauxe, and D. R. Clark, The MatuyamaBrunhes transition recorded from Lake Tecopa sediments
quencefrom the islandof Huahine,J. Geophys.Res.,95,
2713-2726, 1990.
(California), Earth Planet. Sci. Left., 87, 463-472, 1988.
Roperch,P., N. Bonhommet,and S. Levi, Paleointensity
of Valet, J.P., P. Tucholka, V. Courtillot, and L. Meynadier,
the Earth's magneticfield during the LaschampexcurPalcomagnetic constraints on the geometry of the geosionand its geomagnetic
implications,Earth Planet. Sci.
magnetic field during reversals, Nature, 356, 400-407,
Left., 88,209-219, 1988.
1992.
Runcorn,S. K., Polarpath in geomagnetic
reversals,Nature,
Watkins, N. D., and J. Nougier, Excursionsand secularvari356, 654-656, 1992.
ation of the Brunhes epoch geomagneticfield in Indian
Saemundsson,
K., L. Kristj&nsson,I. McDougall, and N.
Ocean region, J. Geophys.Res., 78, 6060-6068, 1973.
D. Watkins, K-At dating, geologicaland paleomagnetic Watkins, N. D., L. Kristj&nsson,and I. McDougall, A destudy oœa 5-km lava succession
in northernIceland, J.
tailed palcomagnetic survey of the type location for
Geophys. Res., 85, 3628-3646, 1980.
the Gilsa geomagneticpolarity event, Earth Planet. Sci.
Sanvet,M., A palaeomagnetic
studyof quaternaryvolcanic
Left., 27, 436-444, 1975.
rocksfrom Turkey, Phys. Earth Planet. Int., 1,403-421, Watkins, N. D., I. McDougall, and L. Kristj&nsson,Upper
1968.
Miocene and Pliocene geoxnagneticsecular variation in
Sherwood,G. J., Paleomagnetism
and magnetostratigraphy the Borgarfjhrdur area of western Iceland, Geophys. J.
of Miocene volcanicsin eastern Otago and Banks PeninR. Astron. Soc., 49,609-632, 1977.
sula, New Zealand,N. Z. J. Geol. Geophys.,31,207-224, Zhu, R., C. Liu, H. Wu, and B. Zhu, Transitional field be1988.
haviour for the Matuyama-Brunhes, Sci. in China, Ser.
Souther,J. G., and D. T. A. Symons,Stratigraphyand paB, Engl. trans/., 34, 1252-1257, 1991.
leomagnetism
of Mount Edzizavolcaniccomplex,northwestern British Columbia, Pap. Geol. Surv. Can., 73-32,
1-48, 1974.
J. J. Love, Department of Earth Sciences,University of Leeds,
Stevenson,D. J., Planetary magnetic fields, Rep. Prog. LS2 9JT, England, U.K. (e-•nail: [email protected])
Phys., 46, 555-620, 1983.
Tanaka,H., M. Kono, and S. Kaneko,Paleosecular
variation (Received June 18, 1997; revised November 8, 1997; accepted
of direction and intensity front two Pliocene-Pleistocene December 18, 1997.)